KV caching

Caching of key-value pairs to avoid recomputation; also provides a mechanism for introspection of earlier computations.

Neighborhood — ranked by edge-count

paper

concept

Introspection
implements
The ability of a model to observe its own past internal states or computations; claimed to be architecturally permitted by transformers.

artifact

Janus Information Flow Transformers (Twitter thread, Sept 2025)
cites
Original thread by janus explaining transformer information highways and introspection capabilities, posted on X.

cosine ≥ 0.65 · no typed edge

Entities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.

KV caching overcomes statelessness and provides a mechanism for introspection of computations at earlier token positions.claim0.793
Janus's claim about KV caching as an introspection mechanism.
KV State Persistenceconcept0.724
The key-value cache from steered tokens is retained during no-steering continuation, allowing causal effect of steering to propagate
K/V Streamconcept0.688
Proposed pathway flowing across positions at each layer; carries key, value, and attention-weighted information horizontally.
Compression sweepconcept0.687
A process by which vasomotion collapses ambivalent neural patterns into durable definite states, reducing complexity.
Cookingconcept0.681
KV state persistence across steered tokensconcept0.668
The causal steering experiment persists KV state over steered tokens so downstream effects can be observed without continued steering
Activation Cappingmethod0.667
Clamping activations along the Assistant Axis to remain above a minimum threshold (25th percentile), introduced as a stabilization method
Contrast-Consistent Search (CCS)concept0.663
Unsupervised probe by Burns et al. to predict latent truth representations; cited as related but limited in generalization